National Science Foundation Engineering Research Center. Bingcai Zhang BAE Systems San Diego, CA

Similar documents
3-D OBJECT RECOGNITION FROM POINT CLOUD DATA

GPU-accelerated 3-D point cloud generation from stereo images

AUTOMATIC TERRAIN EXTRACTION USING MULTIPLE IMAGE PAIR AND BACK MATCHING INTRODUCTION

Lidar and GIS: Applications and Examples. Dan Hedges Clayton Crawford

Presented at the FIG Congress 2018, May 6-11, 2018 in Istanbul, Turkey

CO-REGISTERING AND NORMALIZING STEREO-BASED ELEVATION DATA TO SUPPORT BUILDING DETECTION IN VHR IMAGES

A COMPARISON OF LIDAR TERRAIN DATA WITH AUTOCORRELATED DSM EXTRACTED FROM DIGITALLY ACQUIRED HIGH OVERLAP PHOTOGRAPHY BACKGROUND

Automated Extraction of Buildings from Aerial LiDAR Point Cloud and Digital Imaging Datasets for 3D Cadastre - Preliminary Results

N.J.P.L.S. An Introduction to LiDAR Concepts and Applications

Training i Course Remote Sensing Basic Theory & Image Processing Methods September 2011

EVALUATION OF WORLDVIEW-1 STEREO SCENES AND RELATED 3D PRODUCTS

2010 LiDAR Project. GIS User Group Meeting June 30, 2010

Reality Check: Processing LiDAR Data. A story of data, more data and some more data

Should Contours Be Generated from Lidar Data, and Are Breaklines Required? Lidar data provides the most

Terrain Modeling and Mapping for Telecom Network Installation Using Scanning Technology. Maziana Muhamad

DIGITAL SURFACE MODELS OF CITY AREAS BY VERY HIGH RESOLUTION SPACE IMAGERY

[Youn *, 5(11): November 2018] ISSN DOI /zenodo Impact Factor

2017 PROGRAM OVERVIEW. Geospatial Intelligence for Better Outcomes

UTILIZACIÓN DE DATOS LIDAR Y SU INTEGRACIÓN CON SISTEMAS DE INFORMACIÓN GEOGRÁFICA

VALIDATION OF A NEW 30 METER GROUND SAMPLED GLOBAL DEM USING ICESAT LIDARA ELEVATION REFERENCE DATA

Alaska Department of Transportation Roads to Resources Project LiDAR & Imagery Quality Assurance Report Juneau Access South Corridor

Esri International User Conference. July San Diego Convention Center. Lidar Solutions. Clayton Crawford

AIRBORNE GEIGER MODE LIDAR - LATEST ADVANCEMENTS IN REMOTE SENSING APPLICATIONS RANDY RHOADS

New Features in SOCET SET Stewart Walker, San Diego, USA

AUTOMATED 3-D FEATURE EXTRACTION FROM TERRESTRIAL AND AIRBORNE LIDAR

HEURISTIC FILTERING AND 3D FEATURE EXTRACTION FROM LIDAR DATA

3D CITY MODELLING WITH CYBERCITY-MODELER

TrueOrtho with 3D Feature Extraction

Files Used in this Tutorial

Topographic Lidar Data Employed to Map, Preserve U.S. History

AUTOMATIC BUILDING DETECTION FROM LIDAR POINT CLOUD DATA

1. Introduction. A CASE STUDY Dense Image Matching Using Oblique Imagery Towards All-in- One Photogrammetry

SPOT-1 stereo images taken from different orbits with one month difference

I. Project Title Light Detection and Ranging (LIDAR) Processing

BUILDING DETECTION AND STRUCTURE LINE EXTRACTION FROM AIRBORNE LIDAR DATA

GEOGRAPHIC INFORMATION SYSTEMS Lecture 25: 3D Analyst

Digital Elevation Models (DEM)

Algorithms for GIS csci3225

LiDAR Applications. Examples of LiDAR applications. forestry hydrology geology urban applications

What s New in Imagery in ArcGIS. Presented by: Christopher Patterson Date: October 18, 2017

AUTOMATIC EXTRACTION OF BUILDING OUTLINE FROM HIGH RESOLUTION AERIAL IMAGERY

Improving wide-area DEMs through data fusion - chances and limits

Bonemapping: A LiDAR Processing and Visualization Approach and Its Applications

Light Detection and Ranging (LiDAR)

Visualizing 2D Data in a 3D World

Representing Geography

An Introduction to Lidar & Forestry May 2013

a Geo-Odyssey of UAS LiDAR Mapping Henno Morkel UAS Segment Specialist DroneCon 17 May 2018

Creating raster DEMs and DSMs from large lidar point collections. Summary. Coming up with a plan. Using the Point To Raster geoprocessing tool

3D Analyst Visualization with ArcGlobe. Brady Hoak, ESRI DC

DEVELOPMENT OF ORIENTATION AND DEM/ORTHOIMAGE GENERATION PROGRAM FOR ALOS PRISM

Outline of Presentation. Introduction to Overwatch Geospatial Software Feature Analyst and LIDAR Analyst Software

Overview of the Trimble TX5 Laser Scanner

DENSE 3D POINT CLOUD GENERATION FROM UAV IMAGES FROM IMAGE MATCHING AND GLOBAL OPTIMAZATION

EVALUATION OF WORLDVIEW-1 STEREO SCENES

MODULE 1 BASIC LIDAR TECHNIQUES

FILTERING OF DIGITAL ELEVATION MODELS

DIGITAL SURFACE MODELS IN BUILD UP AREAS BASED ON VERY HIGH RESOLUTION SPACE IMAGES

An Introduction to Using Lidar with ArcGIS and 3D Analyst

THE USE OF ANISOTROPIC HEIGHT TEXTURE MEASURES FOR THE SEGMENTATION OF AIRBORNE LASER SCANNER DATA

BUILDING EXTRACTION FROM LIDAR USING EDGE DETECTION

EVOLUTION OF POINT CLOUD

New Requirements for the Relief in the Topographic Databases of the Institut Cartogràfic de Catalunya

REMOTE SENSING LiDAR & PHOTOGRAMMETRY 19 May 2017

Digital photogrammetry project with very high-resolution stereo pairs acquired by DigitalGlobe, Inc. satellite Worldview-2

8 Geographers Tools: Automated Mapping. Digitizing a Map IMPORTANT 2/19/19. v Tues., Feb. 26, 2019.

DSM GENERATION FROM EARLY ALOS/PRISM DATA USING SAT-PP

Digital Elevation Models (DEMs)

The Feature Analyst Extension for ERDAS IMAGINE

Comparison of LiDAR and Stereo Photogrammetric Point Clouds for Change Detection Paul L Basgall a, Fred A Kruse b, Richard C Olsen b

SimActive and PhaseOne Workflow case study. By François Riendeau and Dr. Yuri Raizman Revision 1.0

8 Geographers Tools: Automated Mapping. Digitizing a Map 2/19/19 IMPORTANT. Revising a Digitized Map. The Digitized Map. vtues., Feb. 26, 2019.

Photogrammetric Performance of an Ultra Light Weight Swinglet UAV

CORRECTING DEM EXTRACTED FROM ASTER STEREO IMAGES BY COMBINING CARTOGRAPHIC DEM

A DATA DRIVEN METHOD FOR FLAT ROOF BUILDING RECONSTRUCTION FROM LiDAR POINT CLOUDS

SOME stereo image-matching methods require a user-selected

POSITIONING A PIXEL IN A COORDINATE SYSTEM

Exelis Visual Information Solutions Capability Overview Presented to NetHope October 8, Brian Farr Academic & NGO Program Manager

Geometric Accuracy Evaluation, DEM Generation and Validation for SPOT-5 Level 1B Stereo Scene

DENSE IMAGE MATCHING FOR MARS EXPRESS HRSC IMAGERY BASED ON PRECISE POINT PREDICTION METHOD

PRODUCTION SYSTEM FOR AUTONOMOUS 3-DIMENSIONAL MODELING WITH LIDAR, IFSAR, AND PHOTOGRAMMETRIC DSM DATA INTRODUCTION

EVALUATION OF ZY-3 FOR DSM AND ORTHO IMAGE GENERATION

TRIMBLE BUSINESS CENTER PHOTOGRAMMETRY MODULE

Surface and Terrain Models

Photogrammetry for forest inventory.

Vegetation height maps derived from digital elevation models the next innovation in the production of orienteering maps?

LIDAR MAPPING FACT SHEET

Generating 50cm elevation contours from space PhotoSat s s new stereo satellite elevation processing system

FREE TUTORING. Digitizing a Map. 8 Geographers Tools: Automated Mapping. The Digitized Map. Revising a Digitized Map 9/28/2018. Next class: First Exam

Assimilation of Break line and LiDAR Data within ESRI s Terrain Data Structure (TDS) for creating a Multi-Resolution Terrain Model

Dense DSM Generation Using the GPU

Technical Considerations and Best Practices in Imagery and LiDAR Project Procurement

Point Cloud Classification

Municipal Projects in Cambridge Using a LiDAR Dataset. NEURISA Day 2012 Sturbridge, MA

Generate Digital Elevation Models Using Laser Altimetry (LIDAR) Data

Producing Ortho Imagery In ArcGIS. Hong Xu, Mingzhen Chen, Ringu Nalankal

Experiences with High density image matching (idsm) at the LVG Bavaria and other NMA in Germany Wolfgang Stößel Photogrammetry and Remote Sensing

NEXTMap World 30 Digital Surface Model

ENVI 5 & E3De. The Next Generation of Image Analysis

UP TO DATE DSM GENERATION USING HIGH RESOLUTION SATELLITE IMAGE DATA

Transcription:

Bingcai Zhang BAE Systems San Diego, CA 92127 Bingcai.zhang@BAESystems.com

Introduction It is a trivial task for a five-year-old child to recognize and name an object such as a car, house or building. However, it is a challenging software problem to identify and label these same objects automatically in a digital image. Geospatial information technology such as digital photogrammetry can answer the where question accurately. The next breakthrough may be the what question, which is to identify and label objects automatically in digital imagery. Automatic 3-D building extraction from digital imagery is considered the Holy Grail in photogrammetry. It is very difficult to automatically extract buildings from images using only their radiometric properties. 2

Introduction continued In the past three decades, many algorithms have been developed to extract 3-D buildings from a very specific set of digital images Until now, there has not been a commercial software package that can reliably do this We show the results of automatically identifying objects such as houses and buildings, including the relationships between DSM accuracy and post spacing, and size requirements for automatically extracting and labeling objects 3

Radiometry vs. 3-D shapes Six different building colors and patterns Very difficult to extract buildings based on radiometric properties only Terrain shaded relief of digital surface model generated by NGATE The locations and approximate shapes of the buildings are obvious 4

Next-Generation Photogrammetry Automation System The two most important technologies in modern photogrammetry are sensor modeling and automation. Sensor modeling provides accurate measurements and automation for terrain generation, while 3-D building extraction increases productivity. 5

Technical approach Automatic transformation from LIDAR point cloud to bare-earth model Bare-Earth Profile Bare-Earth Morphology Bare-Earth Histogram Bare-Earth Dense Tree Canopy Identify and group 3-D object points into regions Separate buildings and houses from trees Trace region boundaries Regularize and simplify boundary polygons Construct complex roofs 6

Sample results: Case study one LIDAR data with a post spacing of 0.2 meters was converted to a GRID format with a post spacing of 0.1 meters The following building parameters were used: 1. Minimum height 2 meters 2. Minimum width 5 meters 3. Maximum width 200 meters 4. Roof detail 0.4 meters 5. Enforce building squaring on AFE transforms a GRID DSM into a GRID DEM using parameters 1, 2 and 3 as the first step 59 buildings and 13 trees extracted 7

Sample results: Case study one continued 8

Sample results: Case study one continued A complex building, with more than 50 sides, automatically extracted by AFE from LIDAR 9

Sample results: Case study one continued To determine the accuracy of automatically extracted building boundaries, we compare segment deltas of the same building boundary extracted by a human operator using photogrammetric stereo images. The stereo images have a pixel resolution of GSD 0.07 meters. Therefore, we can consider the manually extracted building boundary as ground truth for our accuracy analysis. The RMSE is about 0.2 meters or one post spacing. 3-D slant 2-D XY Elevation length delta length delta delta Max 0.910 0.830 0.900 Mean 0.276 0.215 0.080 RMSE 0.394 0.236 0.222 10

Sample Results: Case study two LIDAR data provided by USC s Integrated Media Systems Center (IMSC) with average post spacing of 0.4 meters 138 million posts covering a relatively flat area with many trees The following building parameters were used: 1. Minimum height 2 meters 2. Minimum width 3 meters 3. Maximum width 300 meters 4. Roof detail of 0.4 meters 5. Enforce building squaring on 2464 buildings and houses extracted 5164 trees extracted. 1.2 hours to complete (four 3 GHZ CPUs with debugging code) 11

Sample Results: Case study two continued Terrain shaded relief of 24.8 square kilometers 12

Sample Results: Case study two continued 13

Sample results: Case study two continued 14

Sample Results: Case study two continued 15

Sample results: Case study two continued 16

3-D flythrough 17

Potential applications Data fusion between LIDAR and other types of data such as EO images Registering existing 3-D site models from EO images to LIDAR 3-D site models When LIDAR is much more accurate, this improves the EO sensor parameters Robotics and UAV/UGV applications Make it real-time to navigate a robot Fighting vehicle (navigate and recognize target) 18

Conclusions Automatically identifying, extracting, and labeling 3-D buildings from dense digital surface models by using their invariant 3-D properties can be achieved to a production level capability. Unlike automatic terrain generation systems, which are very mature and have been used widely for two decades, the commercial production level automatic 3-D building extraction system is in its infancy. Our research and development indicates that we can automatically label 3- D buildings by applying advanced stereo image matching technology and LIDAR, automatic bare-earth transformation from DSM, and 3-D buildings invariant 3-D properties. For precise GIS cartographic mapping, manual digitizing should still be used. For fast and affordable 3-D modeling, simulation, and visualization, where accuracy is not as important as affordability and speed, automation technology, such as AFE, may soon gain user acceptance. 19

Thanks!